Neuron Explorer

Qwen3-0.6B

dollar sign before a money amount

A neuron in layer 24 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron reacts to the dollar sign that comes right before a sum of money, such as '$19 billion'. Turning it up did not make the written text talk about money in any clear way.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    maximum of $22.5 million and a minimum of $13.7 million a

  2. #2 peak

    ⏎$3 million grant from the Bill & Melinda Gates Foundation will fund development⏎by

  3. #3 peak

    Dakota Game, Fish and Parks Department Authors are James R. Johnson and Gary E. Larson⏎$19.95 and Free Shipping⏎As Done By:⏎AgBio Communications

  4. #4 peak

    .2$.⏎You could, of course, reach for a calculator here, but I want to show you how to do it without. You can work out $2.4 \times 100$ by moving the dot two spaces

  5. #5 peak

    so I can set my table up like this:⏎grams ice cream⏎My Table of Joy sum is “multiply diagonally and divide by the other number”, or $

  6. #6 peak

    up to $60 per car. A proposed bridge would run from Vancouver Island to the Vancouver

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: how many **opedsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsopsops

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    Argentina east of the Andes, including the islands of Grenada, Trinidad, Tobago, and Margarita. The range of the nine-banded armadillo has undergone rapid expansion into the southern United States since the late 18

  2. #2 peak

    radically separated. As one kills, one can no longer be killed. The enemy can henceforth be eradicated from behind the safety of a computer screen, thousands of miles away from the

  3. #3 peak

    Persons worksheet. If you uncover all eight names, the FBI (Floss Bureau of Investigation) will reward you with the title of Master Word Sleuth.⏎Need the answers? They're here.<|endoftext|>Article I: Pro

  4. #4 peak

    scales, and the tail is encased in a series of bony rings. Coloration of nine-banded armadillos is generally grayish brown, with yellowish-white scales along the side of

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L20:61 strongly suppresses
  2. L22:271 strongly suppresses
  3. L16:39 excites
  4. L21:195 excites
  5. L4:216 excites
  6. L0:141 suppresses the word 'of' in 'middle of'
  7. L2:761 suppresses
  8. L20:863 excites
  9. L16:10 suppresses
  10. L18:48 suppresses
  11. L16:41 excites
  12. L23:109 excites
  13. L19:39 excites
  14. L16:3 suppresses
  15. L16:0 suppresses
  16. L23:474 excites
  17. L1:259 suppresses
  18. L9:7 excites
  19. L21:625 excites
  20. L21:1781 suppresses
  21. L16:37 suppresses
  22. L17:38 suppresses
  23. L16:6 excites
  24. L16:262 suppresses

Feeds later layers

  1. L27:2017 strongly suppresses
  2. L25:814 suppresses
  3. L26:2232 weakly suppresses
  4. L27:744 weakly suppresses
  5. L26:817 weakly excites
  6. L26:56 weakly excites
  7. L27:1396 weakly suppresses
  8. L26:153 weakly suppresses
  9. L25:714 weakly suppresses
  10. L27:166 weakly suppresses
  11. L25:886 weakly suppresses
  12. L26:1977 weakly suppresses
  13. L27:1144 weakly suppresses
  14. L26:2881 weakly suppresses
  15. L27:1786 weakly suppresses
  16. L25:1141 weakly suppresses
  17. L26:493 weakly excites
  18. L26:1808 weakly excites
  19. L27:2086 weakly excites
  20. L25:2898 weakly suppresses
  21. L25:231 weakly excites
  22. L27:1556 weakly suppresses
  23. L25:2187 weakly excites
  24. L25:2377 weakly suppresses

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -35.25 and -33.371 values between -18.28 and -16.401 values between -12.63 and -10.741 values between -10.74 and -8.861 values between -8.86 and -6.971 values between -6.97 and -5.091 values between -5.09 and -3.2028 values between -3.20 and -1.323773 values between -1.32 and 0.5749 values between 0.57 and 2.4528 values between 2.45 and 4.349 values between 4.34 and 6.227 values between 6.22 and 8.111 values between 8.11 and 9.991 values between 9.99 and 11.882 values between 11.88 and 13.761 values between 15.65 and 17.531 values between 28.85 and 30.731 values between 38.27 and 40.16offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. 他也
  2. successes
  3. AAC
  4. cohol
  5. er
  6. 终于
  7. 他已经
  8. 他还
  9. 決め
  10. iating

Suppresses

  1. \
  2. \base
  3. \C
  4. \D
  5. \S
  6. \$
  7. \M
  8. \data
  9. \V
  10. \User